v0.7.1 fix(jobworker): skip error reporting for jobs cancelled during shutdown

[ungerik]

Summary

A job dispatch cancelled during shutdown was still logged as ERROR and sent to
the OnError callback, even though the job is simply reset and retried (no retry
attempt consumed). This is expected shutdown noise, not a failure. This PR guards
both reporting points so a context.Canceled interruption is handled quietly.

What changed

• jobworker/dojob.go — in DoJob, the OnError call and the "Job error"
  ERROR/WARN log are now skipped when the worker's error is (or wraps)
  context.Canceled. The error is still returned and job.ErrorMsg/ErrorData
  are still set, so doJobAndSaveResultInDB still detects the cancellation and
  resets/retries the job exactly as before.
• jobworker/workerthreads.go — the worker dispatch loop skips OnError and
  the "Error while dispatching the job" ERROR log when
  doJobAndSaveResultInDB returns a context.Canceled (the shutdown reset path
  returns ctx.Err()).
• jobworker/dojob_internal_test.go — new DB-free test: OnError is skipped
  for a direct and a wrapped context.Canceled, still fires for a genuine
  failure, and the cancellation still propagates out of DoJob.

The OnError contract ("called for every error that would also be logged") is
preserved: the log and the callback are skipped together.

Why it's correct

• errors.Is(err, context.Canceled) traverses both errs.WrapWithFuncParams
  (via withCallStack.Unwrap) and errors.Join, so the guard fires on the
  wrapped return values (verified by the new test).
• Per-job JobTimeout (context.DeadlineExceeded) is deliberately not
  suppressed — those jobs genuinely ran too long and still consume a retry.

Known consideration (not a blocker)

The guard suppresses any error whose chain contains context.Canceled, including
the rare case where a worker returns a cancel-wrapped error while the worker
thread's context is still alive (not a real shutdown). Such a job is reset and
retried without consuming an attempt — pre-existing behavior — but now with no
log/OnError signal. This matches the intended "a cancelled dispatch is always an
expected shutdown interruption" framing. If we later want to scope it strictly to
shutdown, gate the workerthreads.go guard on ctx.Err() != nil instead.

Testing

./scripts/run-tests.sh — go vet / revive / gosec clean, all packages pass
(jobqueue, jobworker, jobworkerdb, tests), including the new guard test.

Release

Bumps VERSION to v0.7.1 and adds the CHANGELOG entry.